Influence of accelerator pedal force feedback on truck drivers' speed control
Repository URI
Repository DOI
Change log
Authors
Abstract
The UK government has set clear targets for 80% reductions (compared with 1990 levels) in greenhouse gas emissions by 2050 and pressure is increasing on the road transport industry to reduce the fuel consumption and harmful exhaust emissions of Heavy Goods Vehicles (HGVs). Vehicle manufacturers and operators alike are having to investigate and find new ways of making reductions. It is thought that improving driver behaviour offers significant potential for these reductions in fuel consumption and emissions. This thesis considers the use of Active Accelerator Pedals (AAPs) and the potential for improved driver performance that they may offer by providing pedal force feedback to the driver.
In order to develop understanding of the interactions between the human driver and accelerator pedal, two near identical tractor units, operated by Turners of Soham Ltd, were fitted with the a data logger. Data was collected and stored over a period of four months as they operated on the road. This data provided the basis for a vehicle model to be developed using real-world conditions, rather than strictly controlled test track conditions. Analysis of the behaviour of the two drivers also identified differences is styles, and explained 7% fuel consumption differences between the two drivers when negotiating roundabouts.
A new mathematical model of the human driver’s longitudinal control was also developed to include the driver’s cognitive control of the accelerator pedal. Model Predictive Control theory, commonly used for modelling the driver’s steering control, was used and different driving styles were replicated by varying the weightings in a cost function, and a series of driving simulator experiments were performed to validate the model. Nine human drivers, two of which were professionals, performed two driving scenarios (drive cycle and car-following). The driver model was fitted to each driver individually to mathematically express the differences in their styles. The simulated RMS pedal forces from the fitted driver models lay within 20% of the measured simulator values.
The driver model was also extended to include the interactions between a human driver and an AAP using mathematical game theory. Three frameworks were proposed: decentralised, cooperative and one-sided cooperative, but, as the cooperative framework would have been very difficult to implement experimentally, it was only considered theoretically. The same nine human drivers were presented with drive cycle and car-following scenarios whilst being assisted by pedal feedback to validate the model. Both decentralised and one-sided cooperative frameworks were applied to the fitting and compared. In the drive cycle scenario, the one-sided cooperative framework output an identical controller to the decentralised framework. In the car-following scenario, the one-sided cooperative framework produced the best fit, suggesting that the human drivers adapted their strategy to reflect the guidance from the AAP. It was noted in both scenarios that the peak pedal displacement decreased by approximately 20% with the presence of pedal force feedback.
Further work is suggested to improve the mass and road gradient data obtained from the data loggers in vehicles in order to reduce the uncertainty in the traction force and fuel rate maps. With a model for the interactions of a human driver with an AAP now in place, the pedal feedback strategy can now be optimised to improve the performance of the human driver.